Skip to main content

3.4 Gradients of a Vector Valued Functions

 We discussed partial derivatives and gradients of functions $f:\mathbb{R}^n \to \mathbb{R}$ mapping to the real numbers.Lets generalize this concept of the gradient to vector-valued functions(vector fields) $f: \mathbb{R}^n \to \mathbb{R}^m$, where $n \ge 1$ and $m \ge 1$.


For a function $f:\mathbb{R}^n \to \mathbb{R}^m$ and a vector $x=[x_1,\ldots,x_n]^T \in \mathbb{R}^n$, the corresponding vector of function values is given as

$f(x)=\begin{bmatrix}

f_1(x)\\

\vdots \\

f_m(x)

\end{bmatrix} \in \mathbb{R}^m$
where each $f_i: \mathbb{R}^n \to \mathbb{R}$ that map onto $\mathbb{R}$.

The partial derivative of a vector-valued function $f:\mathbb{R}^n \to \mathbb{R}^m$ with respect to $x_i \in \mathbb{R}$, $i=1,\ldots,n$, is given as the vector


We know that the gradient of  $f$ with respect to a vector is the row vector of partial derivatives, every partial derivative $\frac{\partial f}{\partial x}$ is a column vector. Therefore we obtain the gradient of $f:\mathbb{R}^n \to \mathbb{R}^m$ with respect to $x \in \mathbb{R}^n$ by collecting this partial derivatives


Definition: Jacobian
The collection of all first-order partial derivatives of a vector-valued function $f: \mathbb{R}^n \to \mathbb{R}^m$ is called the Jacobian. The Jacobian is an $m \times n$  matrix.
$J=\bigtriangledown _x f = \frac{\mathrm{d f(x)} }{\mathrm{d} x}=\begin{bmatrix}
\frac{\partial f(x))}{\partial x_1} & \cdots & \frac{\partial f(x))}{\partial x_n}
\end{bmatrix}$


As a special case of a function $f: \mathbb{R}^n \to \mathbb{R}^1$, which maps a vector $x \in \mathbb{R}^n$ onto a scalar, possesses a Jacobian that is a row vector of dimension $1 \times n $.

Remark: This is called numerator layout of the derivatives. i.e., the derivative $\frac{\mathrm{d f(x)} }{\mathrm{d} x}$ of $f \in \mathbb{R}^m$ with respect to $x \in \mathbb{R}^n$ is an $ m \times n$ matrix, where the elements of $f$ define the rows and the elements of $x$ defines the columns of the corresponding Jacobian. There exists also the denominator layout, which is the transpose of the numerator layout.


Lets consider an example with $x=\left [x_1 \quad x_2 \right]$.We define two vector valued functions $f_1(x)$ and $f_2(x)$.
If we organize both of their gradients into a single matrix, we move from vector calculus into matrix calculus. This matrix, and organization of the gradients of multiple functions with multiple variables, is known as the Jacobian matrix.
Example:

$f(x,y)=3x^2y$
$g(x,y)=2x+y^8$






Example Problems

1.Consider the following functions
A.$f_1(x)=sin(x_1)cos(x_2), x \in \mathbb{R}^2$
B.$f_2(x,y)=x^Ty,\quad x,y \in \mathbb{R}^n$
C.$f_3(x)=xx^T, \quad x \in \mathbb{R}^n$

a) What are the dimensions of $\frac{\partial  f_i}{\partial x}$
b) Compute the Jacobians.

A.
$f_1(x)=sin(x_1)cos(x_2), x \in \mathbb{R}^2$
$\frac{\partial  f_1}{\partial x_1}=cos(x_1)cos(x_2)$
$\frac{\partial  f_1}{\partial x_2}=-sin(x_1)sin(x_2)$

$J=\bigtriangledown _x f_1 =\begin{bmatrix}
\frac{\partial f(x)}{\partial x_1} & \frac{\partial f(x)}{\partial x_2}
\end{bmatrix}$

$J=\begin{bmatrix}
cos(x_1)cos(x_2)& -sin(x_1)sin(x_2)
\end{bmatrix}\quad \in \mathbb{R}^{1 \times 2}$

B.
$f_2(x,y)=x^Ty,\quad x,y \in \mathbb{R}^n$
$\frac{\partial  f_2}{\partial x}=\left[\frac{\partial  f_2}{\partial x_1}\quad \frac{\partial  f_2}{\partial x_2} \cdots \frac{\partial  f_2}{\partial x_n}\right]=\left[y_1 \quad y_2 \quad \cdots y_n \right]=y^T$

$\frac{\partial  f_2}{\partial y}=\left[\frac{\partial  f_2}{\partial y_1}\quad \frac{\partial  f_2}{\partial y_2} \cdots \frac{\partial  f_2}{\partial y_n}\right]=\left[x_1 \quad x_2 \quad \cdots x_n \right]=x^T$

$J=\begin{bmatrix}
\frac{\partial f(x)}{\partial x_1} & \frac{\partial f(x)}{\partial x_2}
\end{bmatrix}$

$J=\begin{bmatrix}
y ^T & x^T
\end{bmatrix}\quad \in \mathbb{R}^{1 \times 2n}$

C.
$f_3(x)=xx^T, \quad x \in \mathbb{R}^n$

2.Differentiate $f$ with respect to $t$ ( university question)
$f(t)=sin(log(t^Tt)) \quad t \in \mathbb{R}^D$

$f(t)=cos(log(t^Tt))\frac{1}{t^Tt}2t^T$

3.Compute the derivatives $\frac{\mathrm{d} f}{\mathrm{d} x}$ of the following functions by using the chain rule. Provide the dimensions of every single partial derivative. Describe your steps in detail.     ( university question)

$f(z)=log(1+z),\quad z=x^Tx\quad x \in \mathbb{R}^D$

4.Compute $\frac{df}{dx}$
$f(z) = sin(z) ; z = Ax + b ; A \in  R^{E\times D}; x \in R^D; b \in R^E$
where $sin(.)$ is applied to every element of $z$. ( university question)



Comments

Popular posts from this blog

Mathematics for Machine Learning- CST 284 - KTU Minor Notes - Dr Binu V P

  Introduction About Me Syllabus Course Outcomes and Model Question Paper University Question Papers and Evaluation Scheme -Mathematics for Machine learning CST 284 KTU Overview of Machine Learning What is Machine Learning (video) Learn the Seven Steps in Machine Learning (video) Linear Algebra in Machine Learning Module I- Linear Algebra 1.Geometry of Linear Equations (video-Gilbert Strang) 2.Elimination with Matrices (video-Gilbert Strang) 3.Solving System of equations using Gauss Elimination Method 4.Row Echelon form and Reduced Row Echelon Form -Python Code 5.Solving system of equations Python code 6. Practice problems Gauss Elimination ( contact) 7.Finding Inverse using Gauss Jordan Elimination  (video) 8.Finding Inverse using Gauss Jordan Elimination-Python code Vectors in Machine Learning- Basics 9.Vector spaces and sub spaces 10.Linear Independence 11.Linear Independence, Basis and Dimension (video) 12.Generating set basis and span 13.Rank of a Matrix 14.Linear Mapping...

4.3 Sum Rule, Product Rule, and Bayes’ Theorem

 We think of probability theory as an extension to logical reasoning Probabilistic modeling  provides a principled foundation for designing machine learning methods. Once we have defined probability distributions corresponding to the uncertainties of the data and our problem, it turns out that there are only two fundamental rules, the sum rule and the product rule. Let $p(x,y)$ is the joint distribution of the two random variables $x, y$. The distributions $p(x)$ and $p(y)$ are the corresponding marginal distributions, and $p(y |x)$ is the conditional distribution of $y$ given $x$. Sum Rule The addition rule states the probability of two events is the sum of the probability that either will happen minus the probability that both will happen. The addition rule is: $P(A∪B)=P(A)+P(B)−P(A∩B)$ Suppose $A$ and $B$ are disjoint, their intersection is empty. Then the probability of their intersection is zero. In symbols:  $P(A∩B)=0$  The addition law then simplifies to: $P(...

5.1 Optimization using Gradient Descent

Since machine learning algorithms are implemented on a computer, the mathematical formulations are expressed as numerical optimization methods.Training a machine learning model often boils down to finding a good set of parameters. The notion of “good” is determined by the objective function or the probabilistic model. Given an objective function, finding the best value is done using optimization algorithms. There are two main branches of continuous optimization constrained and unconstrained. By convention, most objective functions in machine learning are intended to be minimized, that is, the best value is the minimum value. Intuitively finding the best value is like finding the valleys of the objective function, and the gradients point us uphill. The idea is to move downhill (opposite to the gradient) and hope to find the deepest point. For unconstrained optimization, this is the only concept we need,but there are several design choices. For constrained optimization, we need to intr...